Lead Conversion Prediction (MIT IDSS – Capstone Project)

As part of an MIT-IDSS capstone project, a predictive analytics system was built to identify leads most likely to convert into paying customers for the fictional EdTech company ExtraaLearn.

The workflow began with data cleaning — removing duplicates, reconciling inconsistencies, validating ranges, and addressing anomalies — followed by univariate and bivariate exploratory analysis.

Feature engineering was performed to create engagement scores, interaction ratios, and normalised lead scores that better represent user behaviour, then multiple models were evaluated:

• Decision Trees, Random Forest, Gradient Boosting
• Logistic Regression, SVM

This was followed by hyperparameter tuning, threshold optimisation, and precision–recall tradeoff analysis. From the best-performing model, feature-importance insights were extracted and lead profiles were developed to support targeted marketing.

Tech Used: Python, Google Colab | Scikit-learn, Pandas and Seaborn (main libraries)

Notebook Preview

Due to MIT-IDSS copyright restrictions, the notebook cannot be shared publicly. I am happy to discuss further details upon request.